Paramed

Paramed is a port of the eponyous Stata program to conduct causal mediation analysis using parametric regression models proposed by Tyler VanderWeele. The original program is available from https://cdn1.sph.harvard.edu/wp-content/uploads/sites/603/2018/03/MediationPsychMethods.zip.

Installation

  Pkg.clone("https://github.com/mwsohn/Paramed.jl")

Syntax

paramed(yvar::Symbol,avar::Symbol, mvar::Symbol, a0::Int, a1::Int, m::Int, df::DataFrame; interaction::Bool = true,controlvars = [], logfile::IOStream = nothing)

Options

Positional parameters

The following parameters must be provided in the order in which they are listed below.

• yvar - dependent variable (Symbol)
• avar - analytic variable (Symbol)
• mvar - mediator variable (Symbol)
• a0 - natural level of the treatment (exposure)
• a1 - alterantive treatment (exposure) level
• m - the level of mediator at which the controlled direct effect is to be estimated. If there is no treatment-mediator interaction, the controlled direct effect is the same at all levels and so an arbitary value can be chosen.
• df - DataFrame containing the analytic data

Keyword parameters

• interaction - A Boolean indicating whether treatment-mediation analysis is to be included (default: true)
• controlvars - An array of Symbols for control variables